Generative AI

Benefits of Gen AI

According to Microsoft Copilot: Artificial Intelligence (AI) has the potential to bring about numerous benefits to society across various domains. Here are five ways AI can positively impact society:

Healthcare Advancements

AI can assist in early disease detection, diagnosis, and treatment planning. Machine learning algorithms can analyze vast amounts of medical data, such as images and patient records, to identify patterns and make accurate predictions. This can lead to quicker and more precise diagnoses, personalized treatment plans, and improved patient outcomes. One AI tool can detect early-stage Alzheimer’s through voice analysis (Canary Speech, based in Utah), and AI-supported mammogram screening increases early detection by 20%. There is also a case where a mother used ChatGPT to correctly diagnose her child’s medical situation when 17 doctors over three years failed.

Environmental Conservation

AI can be used to monitor and manage environmental resources more effectively. For instance, AI-equipped sensors and drones can help track changes in ecosystems, predict natural disasters, and manage resources like water and energy more efficiently. This technology can contribute to sustainable practices and help mitigate the effects of climate change.

Efficient Transportation

AI-powered systems can optimize traffic flow, manage public transportation networks, and even facilitate the development of autonomous vehicles. These technologies can lead to reduced congestion, shorter commute times, and increased safety on the roads. Additionally, AI can contribute to route planning, reducing fuel consumption and greenhouse gas emissions.

Improved Accessibility

AI can enhance accessibility for individuals with disabilities. Speech recognition, natural language processing, and computer vision technologies can assist those with hearing, speech, or visual impairments by enabling them to interact with digital devices and services more easily. AI can also help bridge language barriers, making information and communication more inclusive.

Manufacturing and Customer Service

AI can optimize production processes, design new products, and improve supply chain management, leading to increased efficiency and innovation. AI-powered chatbots and virtual assistants can offer 24/7 support, improving customer satisfaction and efficiency in handling inquiries and issues.

Finance

AI can help with fraud detection, risk management, and personalized financial advice, making financial services more secure and tailored to individual needs.

Enhanced Education

AI-powered educational tools can provide personalized learning experiences to students. Adaptive learning platforms can assess individual students’ strengths and weaknesses and tailor educational content accordingly. Virtual tutors and intelligent chatbots can offer real-time help and guidance, improving access to education and supporting both students and educators. ChatGPT does a good job of defining and explaining concepts in an understandable way to all learning levels. GenAI can assist with grammar and proofreading, especially when the student is not working in their native language.

Limitations and Concerns

There are still several limitations with generative AI tools:

• They may provide plausible-sounding but inaccurate responses and most don’t provide citations, so it can be difficult to evaluate the responses.
• The tools learn from human trainers and from consuming the Internet, so more work is needed to eliminate bias.
• Although trained to refuse inappropriate requests and detect false statements, the technology is not always accurate, in part because it may not be up-to-date on the most recent news/developments.
• The tools are expensive to maintain so those that are currently free may not remain free.​ This would increase the digital divide by providing such a powerful tool only to those who could afford it. This is one reason why banning access to AI tools on campus computers may not be a good thing.

Here are some more specific examples:
Bullying & Stereotypes	Bias
Lensa, an AI avatar generator, created highly sexualized images, especially with women, lightened skin tones & anglicized features, full nudes from headshots, sexualization of minors (Read more) Microsoft limits its Bing AI chats after it had some very unsettling conversations.	One study has found that AI detection tools falsely accuse international students of cheating.​ ​AI chatbots are less fluent in languages other than English. It can be difficult to get AI to generate results that are not biased toward Western European perspectives. For example, when asking AI to generate an image of a doctor, most results will depict white men. The AI companies are working to improve this.
Copyright	Influence
Copilot and ChatGPT list creative industries and entertainment as positive impacts of using generative AI. AI can assist in generating music, art, and writing, providing new tools and inspiration for artists, writers, and musicians; and AI can create immersive experiences in gaming, virtual reality, and interactive storytelling, enhancing the entertainment landscape. However, the darker side to this is that AI “learns” largely by scrubbing the web for information. Artists, novelists, and other artists have very little ability to prevent AI from consuming and imitating their works and styles.  Are A.I. Image Generators Violating Copyright Laws? AI image generators scrub the web for existing digital images and artwork to create their digital images.  There have been several lawsuits brought by artists and authors alleging copyright violation because the GenAI tools have consumed their work without permission. For example, from The Verge: “Getty Images has filed a case against Stability AI, alleging that the company copied 12 million images to train its AI model ‘without permission … or compensation.’” Similar issues have arisen from authors as well, where AI models have consumed entire books and can “write a book in x author’s style.” There have been a few recent developments related to copyright: Researchers at the University of Chicago have developed a new technique that allows artists to embed invisible “poison” into their work that misleads A.I. models. The tool, called Nightshade, changes an image’s pixels in a way that humans can’t detect. (read more)​ In 2023 a US court in Washington, DC ruled that AI-generated content without human input cannot be copyrighted.	“Influence” is a pretty generic topic for what is possibly the biggest concern around AI-generated content. Broadly speaking, the Internet especially is already at the point where we all need to learn to question everything we see and read, including images and video. There are many examples of AI-generated, false content that have gone viral online, with most people not realizing the content is completely fabricated. At best, this content is perpetuating lies and fabrications, but at worst it can be extremely harmful. Consider how much damage can be done in a political race with an AI-generated image or video depicting someone doing or saying something that did not actually happen. This kind of thing can be damaging not only to politicians and celebrities, but also to each one of us. Imagine a stalking scenario or an angry student and how they could use AI to damage someone’s character. Another concern around influence is how believable AI bots can be and the very human attachments some people form with them. Google’s AI was convincing enough to persuade a Google engineer to declare it sentient at the cost of his job. At least one, and probably more, young person has died by suicide after falling in love with an AI bot. The tag line for an article titled “The Perfect Girlfriend” by Michal Lev-Ram, is “Flirty, sexy, seductive, supportive. Your AI companion can be whatever you want her to be. And now a growing number of men are turning to bots to ease their loneliness or satisfy their kinks. The choices are endless. The emotions are real.” AI bots can provide companionship for those who are lonely/isolated, but this raises so many concerns, just one of which is the depiction of women.
Environmental Impacts	Deepfakes
This concern is not as directly related to teaching, but it is one that you may feel strongly about yourself and is likely to come up when having conversations with students about GenAI.  For more information, see this article from the Harvard Business Review. At the moment it’s a difficult conversation to have because on the one hand, there’s no question the environemtal impacts are large, but on the other hand, AI is already so integrated into our technological lives that it’s difficult to get away from it. We have to figure out our “hybrid car” approach to AI and how to find the middle ground of responsible use while limiting negative effects.	Deepfakes are related to the information on harmful influence. As Leon Furze states:  A deepfake is an emergent type of synthetic mediathat uses artificial intelligence and machine learning (AI/ML) techniques to create or manipulate audio, video, images, or text. Deepfakes produce highly realistic and convincing content of events or people doing or saying things that never actually occurred. Deepfakes are a serious concern. You can read more about it in this article by Leon Furze.

In addition to the above, in 2023 more than 1,000 technology leaders and researchers urged artificial intelligence labs to pause the development of the most advanced systems, warning in an open letter that A.I. tools present “profound risks to society and humanity.” (see NY Times for more info)

Image Generators

Below you can see the result of the image prompt for “little girl holding a teddy bear in the middle of nowhere.” The images were created by:

• Midjourney: hyperealistic, steep learning curve, available within Discord and associated cost after free trial​
• DALL-E: by OpenAI, better with text in images​
• Adobe Firefly Image 2: doesn’t train on unauthorized art, has “Content Credentials” to show how an image was made and by whom​

Academic Integrity

The graphic below from World Economic Forum addresses the potential risks and benefits of using AI in higher education:

It is possible to embrace using GenAI in the classroom and still be concerned about academic integrity. As instructors, we want to be sure that our students are learning the foundational material needed to be able to use GenAI ethically and effectively. Although there are AI detection tools out there, including Turnitin, it’s important to know that none of the AI detectors have been proven accurate enough to be relied on at this time because there are large numbers of false positive and negative results. In the following tabs, you’ll have an opportunity to consider some AI-resistant (and AI-inclusive) activities you can integrate into your courses, but the best option, if you have concerns about an academic integrity violation related to AI, is to talk to the student.

Student Perspectives and the Workforce

Student Perspectives

“I am a bit of a technophobe; computer science majors seem like superheroes to me… it would be brilliant if, instead of having to figure this all out on my own, AI could be integrated into my education. I’m not asking for a full-fledged academic AI revolution, in which we’re expected to use AI in all our work. I just want to be prepared to navigate the AI-fueled future. Teach me how to streamline my research processes through AI. Explain to me what questions to ask AI chatbots to get the most helpful responses. Show me how I can use these resources to improve my administrative efficiency and my data analysis. Help me receive edits and constructive criticism from AI. Prepare me for the real world, where AI is beginning to touch all areas of work.”​
-A.N., UCLA Student
 

Additionally, when speaking to a number of students at UNCG, another consistent theme that emerged is the fact that the prevalence of AI online has had a detrimental impact on student mental health. All of the students I spoke with said that GenAI has brought them to the point that they no longer believe ANYTHING they see and read online. Although you could argue that it’s better to question everything than to question nothing, the constant questioning of a large part of your reality can and does bring on a lot of anxiety. Students seem to be craving some stability and we have an opportunity to at least help ease this anxiety by teaching them strategies to work with and recognize GenAI content.

Entering the Workforce

The UCLA student’s is of course only one perspective on GenAI use in higher education, but it is a perspective shared by many students. However, even for those students whose primary goal is to submit assignments as quickly and easily as possible, there is at least one thing shared across all student perspectives: the necessity of being prepared for GenAI use in the workforce. 

AI technology is growing at an exponential rate and the academic and professional worlds must stay current. AI-generated content has the potential to significantly impact the workforce and is already used extensively in both the academic and private sectors.

Although AI can be useful in the private sector for things like data interpretation and coding, businesses and employers must worry about the legal implications of using AI, inaccurate information generated by AI, and the risk of confidential information accidentally being shared. GenAI could be effectively used as an aid to research, but could easily be abused and research integrity is already a real problem (see RetractionWatchLinks to an external site.). Ethics is also a concern when using AI, which sometimes reinforces stereotypes.

We need to prepare our students. Most, if not all, of our students will use GenAI in their careers, so we need to make sure they understand all these issues so that they can use AI in the most effective and ethical ways. We can also consider what kind of AI-related skills students might need for employment, such as how to write effective AI prompts. Another concern that is unfortunately very real is that GenAI will replace some jobs (and already has). We need to think about what jobs are most vulnerable to being replaced by AI and how to set our students up with valuable skills so they can be competitive in a world of GenAI. Finally, students (and all citizens) need to understand how to evaluate text and multimedia to determine what is real and what is fake.

The graphic below from World Economic Forum indicates that AI will be the third highest skill priority for 2027. Note that all of the other top 10 skills are subjects that the College of Arts & Sciences already excels at teaching in all of our fields!

It’s a good idea for instructors to:

• Make an effort to stay current on how AI impacts your area, considering both useful applications and how AI might replace certain jobs/skills.​
• Incorporate AI into your classes- your students will use AI in their careers, so prepare them.
• Uncertain where to begin? Ask ChatGPT!

Leveraging AI in Higher Education

Academic Integrity: Talk to the Students

There are a lot of great ways GenAI can be used in higher education, but first we should address the AI elephant in the room: academic integrity. There are times when it is essential that instructors either prevent or know for sure when their students have used GenAI. If you are trying to monitor for plagiarism, it is important to know that the current tools for detecting AI-generated content are not reliable. Turnitin now has AI detection built into its Feedback Studio. However, in testing the results were all over the place. OpenAI, the creator of ChatGPT, released a software tool to identify text generated by artificial intelligence but eventually pulled it because of low accuracy (see also). Also, One study has found that AI detection tools falsely accuse international students of cheating.

In addition to (or perhaps instead of) using detection tools, there are some things we can do now to try to address AI usage both in and out of the classroom:

• Stay informed about AI capabilities and limitations.​
• Engage in dialogue: communicate with students about when it is appropriate to use AI as an aid and when it is not, as well as potential consequences (both academically and in the learning process).​ Discuss with students some real-world examples of AI use and consequences so students can personally connect actions and consequences. Engage students in a conversation about why they might use AI tools.
• As instructors, we can watch for inconsistencies in student submissions. Collect multiple smaller writing samples throughout the semester. Is the quality of a paper submission significantly different from that of discussion posts or short reflections? If you have concerns, you should meet with the student and ask pointed questions about the assignment content.
• Consider how AI is used in your field and the implications for your course(s). Are there times in your course when student learning can benefit from using the tool?​
• Lean into authentic, personalized assessment whenever possible. Projects and presentations that require students to demonstrate their knowledge, especially if they require multiple methods (writing, images, video, charts, etc) are more difficult for students to cheat on. When possible, ask students to incorporate personal experiences, current events, and quotes from YOUR lecture material.
• When the course allows, spend some time discussing AI with your students, how they can detect AI-generated content themselves, and most importantly why it is essential that they take the time to do so rather than just assume that if the content looks authentic, it must be.
• This Inside HigherED article provides some useful suggestions from other instructors.​​
• If the nature of your course requires that you be as certain as possible that students are turning in authentic work at all times, then consider requiring all assignments be completed in-person, in the classroom, and/or use online or in-person proctoring.

Positive Ways AI Can Be Used in Higher Ed

These excellent graphics from Educause illustrate several positive ways that AI can be leveraged in higher education. Flip the cards to learn how AI can help with dreaming, drudgery, design, and development:

Personalized Learning, Adaptability, Instant Feedback, Increased Engagement

There are already many examples of the positive impact of GenAI in education. None of these examples involved replacing the instructor; rather, they focus on personalized learning, instant feedback, adaptability, engagement, and universal design for learning.

Use Case	Impact
Carnegie Learning’s AI-Powered Math Program MATHia	Students achieved significant improvements in math scores compared to those who didn’t. The personalized nature of the AI system helped struggling students catch up and advanced learners to progress further.
IBM Watson’s Tutoring System	Students had higher pass rates and course completion rates compared to students who did not. The AI system provided instant feedback and tailored resources to meet individual student needs.
DreamBox Learning’s Adaptive Math Platform	Schools reported improved math proficiency among their students. The AI system allowed students to work at their own pace and provided additional support when they faced challenges.
Duolingo’s AI-Powered Language Learning App	Students improved their language skills more than students who used traditional language learning methods. The adaptability of the AI system helped learners focus on their specific language weaknesses.
Knewton’s Adaptive Learning Platform	Students achieved higher grades and lower failure rates compared to students using traditional textbooks. The AI system tailored content to match each student’s learning pace and style.

In addition, GenAI can help both students and instructors with increasing efficiency and expanding ideas/brainstorming:

Use Case	Impact
Increasing efficiency	Creating discussion questions, rubrics, assignment prompts, study aids, quiz questions​; Saving time on simple but repetitive tasks, such as student recommendation letter templates;​ Generic comment/feedback generator so you can focus on the personalized comments​; Additional student learning support outside class hours
Expanding ideas and brainstorming	Faculty and students can use AI to role-play or as a debate partner​; GenAI can suggest ideas for improving lesson plans, discussions​; Fleshing out outlines​; Use with advising to help students connect interests to fields of study​; Provide multiple alternative explanations for complex topics​; Functioning as a study aid/study partner​; brainstorm ideas for how to present content in ways to benefit all learners

Developing AI-Resilient Courses

This simply means creating courses that can withstand and adjust to advances in GenAI in both academics and the field of study. This is a three-pronged approach:

1. Evaluating the impacts of AI on your field of study: Evaluating your subject matter broadly to determine how it might need to be modified to account for the impact of AI in the field. What tasks is AI likely to replace? What knowledge and tasks are best suited to human efforts? How can humans and GenAI work together in the field?
2. At the departmental level: Examine the curriculum and make determinations related to AI, including:
   • Determine what you want your students, especially majors, to learn about AI in relation to the field. What information should ALL students get and what information might be topic-specific?
   • Determine which course(s) could offer the fundamental knowledge, which courses have topic-specific opportunities, and which courses need to approach AI-integrated content with extra caution. 
   • Based on the above decisions, develop a few syllabi statements faculty can choose from, depending on how they will approach AI in their courses.
3. In individual courses: Examine when and how to integrate GenAI into the design of the course (when and how you and your students will use GenAI in the course). Consider creating a list of your SLOs and the assignments that measure student learning, then ask: can AI replace this skill? If yes, then how can the assignment be modified to include AI or replaced with more AI-resilient skills?

Policies and Syllabi Statements

The first step to addressing AI in your courses is to have a clear dialogue with your students. Familiarize yourself with UNCG’s AI Policy and include that as well as your own guidelines for use in your syllabi. When crafting use guidelines, be mindful of how you define generative AI: remember that tools in everyday use, such as text autocomplete and Google searches, are AI. Most programs, including Microsoft, now have generative AI integrated. Generally, if not otherwise stated in your syllabus, any AI use should be treated as it would if the help came from another person.

Crafting Assignments with AI in Mind

• Incorporate oral assessments in-person or virtually.​
• Emphasize process more than final product.​
• Have students incorporate personal experience in assignments/make assignments more meaningful to students.​
• Require direct quotes and examples from the course content.​
• Have students incorporate very recent current events.​
• Scaffold assignments using smaller, more frequent writing assignments.
• Mixed mediums: create assignments that require students to submit a combination of materials: text with images or video, presentations, etc.​
• Ask students to show their process: turn in notes, outlines, rough drafts with comments, mind maps, journals, etc​
• Be very clear to your students about AI expectations in the course. Consider incorporating AI into assignments but with clear parameters.​

Higher Ed might need to focus more on “soft skills” like communicating effectively, critical thinking, and problem-solving skills. All things the College excels at; this could be an opportunity to evaluate what employers want and need, and what skills we’re graduating students with.​

Is it Okay to use AI?

Once you and your students have established your policies around AI use, it might be a good first step to share this flowchart (or something similar) created by Aleksandr Tiulkanov, AI and Data Policy Lawyer, ​January 2023​

Appearing in UNESCO’s ​ChatGPT and Artificial Intelligence in Higher Education Quick Start Guide​

Citations

“ChatGPT, the popular AI language model, is quite new. Educational institutions and style guides are still working out their policies on when and how content from the tool can be used and cited in academic writing.”​ -Scribbr​

How Students Can Incorporate AI

In addition to what you’ve already read, these are just a few more ideas of how students can use GenAI:

• Personal Tutor: One of the most beneficial uses of AI in education, generative AI can provide just-in-time assistance, meet the student where they are by providing adaptive learning examples, immediate feedback, study assistant by quizzing the student, explaining complex concepts in multiple ways.
• Universal Learning and Accessibility: translation tools, video captions, text-to-speech, access to real-time, after hours assistance, presenting content in multiple ways/multiple explanations.
• Brainstorming:If students are struggling to come up with ideas for paper topics, experiments, etc, gen AI can provide ideas, role play as a character from history, fiction, etc.​
• Research Assistant: Gen AI find relevant sources on research topics, create full outlines for papers, presentations, workshops/training, suggest topics.​
• Writing Suggestions: Gen AI can provide writing feedback, including suggestions for grammar, conciseness, strength of argument, content improvement and summary, and reading level, language assistance, allow students to check their own work for plagiarism.​
• Marketing & Communication: tweet generation, attention-grabbing headlines, effective summarization; help students explore self-expression through creating images based on feelings, writing samples, etc.

Writing Effective prompts

Crafting effective prompts for AI is essential to get accurate and relevant responses. Prompts act as the guide to producing the desired and correct output. There are a few steps you should take before you begin crafting your prompts.

1. Clarify Learning Objectives: Start by identifying clear learning objectives for what you help to get from the exercise. This will help you stay on task as you navigate AI results and fine-tune your prompts. In particular if you are writing prompts related to course content and assignments, make sure to revisit your SLOs: what do you want students to achieve or demonstrate and does the AI prompt and response contribute to these desired outcomes? It’s easy to get sidetracked with AI results, so you want to try to stay on target with your goals.
2. Choose the best model for the task: Stay informed about the capabilities and limitations of AI models. Choose the model most likely to respond accurately to your prompts, for example you might use ChatGPT or MS CoPilot for text results and Adobe Firefly for image generation.

Jose Antonio Bowen’s Four Pillars of Better Prompt Writing:

Components of a Prompt	Details
Task	Instructions should be specific to desired outcome; use words like apply, transform, reapply; do you want the AI to generate an email, create an announcement, summarize a reading, etc
Format	What kind of output do you want? Is it a script, essay, etc? And how long? Do you want a bulleted list, an email, a bulleted list?
Voice	Write in academic language, medical language, humor, write like Oprah, etc; give the AI a character/persona and a tone
Context	Providing more specific details, such as writing examples. what details about the topic/task will help to generate a more relevant response; if you are teaching a course, you can include more details about the course content; you can provide an example structure of an outline you want it to create; you can ask the model if it needs more information

Example of Applying the Four Pillars:

TASK: You want to send an end of semester course announcement to students.

FORMAT: 3-paragraph end-of-semester course announcement.

VOICE: friendly and encouraging.

CONTEXT: community college, first year experience course; have progressed in research writing.

Complete AI Prompt: Write a three-paragraph end-of-semester announcement to community college students who completed a first-year experience course. Let students know how proud I am of their accomplishments in research skills. Use a friendly and encouraging tone.

Refining Outputs

Asking a poorly formed prompt can still generate a persuasive but inaccurate, and possibly dangerous, response. Once you have your initial prompt, you can refine the output by using the guidance below. Keep in mind that working with GenAI is a bit like having a conversation, so you can build on the previous prompt without having to rewrite it. For example, you can say, “It’s too long. Shorten it to 150 words.”  These guidelines are inspired by the RACCCA Framework (Relevance, Accuracy, Completeness, Clarity, Coherance, Appropriateness):

• Did GenAI create the right format? For example, did it create a bulleted list, a chart, or a table? If not, ask it to regenerate the output in the form of a list or chart.
• Is the tone and voice what you expected? If not, you can ask GenAI to use a friendlier tone or tone down the enthusiasm a bit. You can also provide examples of your writing and ask it to use this example to rewrite the output.
• Would additional background or constraints help to refine the output? For example, adding the type of learner or their reading level can help GenAI to produce an output that is at the appropriate level of challenge.
• Does the output contain bias? Bias occurs due to people’s preferences, prejudices, and stereotypes and can emerge because of how you phrase your prompts. If you ask a question in a leading way, it will produce a biased answer. For example, if your prompt asks GenAI why living in northern states is better than living in southern states, it will produce an output that provides just the benefits of living in the North and the challenges of living in the South. Instead, consider wording your prompt with phrases such as “the pros and cons” or “why or why not” to avoid bias.
• Finally, be sure to check the accuracy of the output. If the output includes references, be sure to check them to ensure they are not hallucinations. Is the content accurate and free from bias? Is the information up to date and accurate? Is it missing any key points that would make it more complete? Is the response appropriate for the audience?

Digital Literacy

It’s more important than ever that we prepare our students to live in a world where technology, and now AI, is integrated into almost all areas of everyday life, both personal and professional. There are many facets to digital literacy with GenAI, including knowing what data is being collected, how it’s being used, and who owns the information. According to the Schwartz Reisman Institute for Technology and Society:

Digital literacy is particularly important in democracies—political systems that rely on citizen knowledge, participation, and choices to govern. Some countries are ahead of the curve. For instance, digital literacy is part of the core school curricula in Finland and Estonia. Students learn to code from a young age and take media and disinformation courses… LLMs pose a serious risk to democracy because they disrupt our ability to access high-quality information, a critical pillar of democratic participation. Basic rights such as the freedom of expression and assembly are hampered when our information is distorted. We need to be discerning consumers of information in order to make decisions to the best of our abilities and participate politically. We need to understand how LLMs (and other AI technologies) generate their answers in order to make use of these powerful tools. We tend to fall prey to automation bias, downgrading human decision-making in favour of the machine. But perhaps that’s because we don’t often think about how the machine works to produce answers. How does a tool like ChatGPT gather and deliver information to me? How can I use the chatbot to spark my creativity instead of making it speak for me? What are the limitations of this tool? How are algorithmic choices biasing the output of these tools? (source)

Higher education is ideally situated to address these issues, in particular because those who already have subject matter knowledge will have the best success with using generative AI because they will be more likely to notice inaccuracies and verify information. At a basic level, we all need skills to understand:​

1. Technical Understanding: Knowing the principles of how AI works and its capabilities; an understanding of how AI will impact the job sector.
2. Ethical Understanding: The ethics of how and when it’s okay to use AI; an understanding of biases, content ownership, and privacy concerns.​
3. Practical Understanding: Knowing how to use the various tools to generate the best results, including effective prompt writing; and strategies for how to detect when something is generated by AI.​

Our students need skills to successfully navigate the internet, safely consume content, and use technology to create products and services that are both ethical and effective. As GenAI tools evolve, it’s increasingly difficult to detect whether some piece of content has been generated by AI. Trying to authenticate every article, image, or video you access can be extremely time-consuming and exhausting. For those who have some awareness of GenAI, especially younger generations like our students, many report constant anxiety from a perpetual state of never believing anything they see or hear (a recent student survey confirmed our UNCG students are feeling this way). On the other hand, a large part of the population continues to take at face value whatever they see, especially if it reinforces what they already believe or want to be true.

GenAI detection tools have so far struggled to keep up as the GenAI tools themselves evolve quickly. Yet even if you embrace AI usage in your course, there are times when you will want to restrict student use of AI tools. Here are some AI detection strategies you can use and share with your students. Please bear in mind that not all of these are not 100% effective; they serve as more of a starting point for a conversation with the student.

Tips & Strategies

Text

AI-generated text is one of the hardest to detect reliably, while also being the most common area of faculty concern. Below are a few tips, but if you have any concerns, the only way to be sure is to talk to the student: see if they can answer questions about what they wrote and why.

• Turnitin: Instructors can enable Turnitin on Assignments (but not quizzes). Turnitin has a plagiarism detection component. Please remember that this tool is not reliable- it generates a significant number of false positives and negatives. Even though you cannot take the results at face value, it may be worth turning it on because it can serve as a deterrent to some students.
• Use the version history tool in word processing to look for large chunks of pasted text with few or no edits.
• Require reference to very current events, because the models are generally a little behind.
• Look for unnatural repetition of phrases.
• Require direct quotes and citations in all writing assignments, even discussion posts.
• Look for more casual, humorous writing. AI-generated content often lacks a unique voice or perspective. The prompt for the example below was “Write a haiku about cats:

Prompt: Write a haiku about cats:

AI-Generated Submission	Original Writing Submission
Whiskers in moonlight,​ Silent paws dance through the night,​ Cats’ grace, pure delight.	In deep sleep hear sound  cat vomit hairball somewhere ​ will find in morning. ​

Prompt: What societal characteristics do you see that helped Athens flourish while Sparta declined?

AI-Generated Submission

Original Writing Submission

Ancient Athens thrived due to its unique democratic system of governance, fostering civic engagement and innovation. In contrast, Sparta struggled because of its rigid and militaristic society, which limited individual freedoms. Athens embraced a culture of intellectual exploration, with philosophers like Socrates and Plato inspiring critical thinking and creativity. Sparta, on the other hand, prioritized military training, stifling intellectual development. Trade and commerce flourished in Athens, thanks to its strategic location and a democratic government that encouraged economic activities. In contrast, Sparta’s focus on conquest and military prowess hindered economic growth, as it prioritized self-sufficiency … Sparta’s strict hierarchical system led to discontent among its helot population, destabilizing the state. Ultimately, Athens thrived by embracing democratic values, intellectual pursuits, and trade, while Sparta struggled due to its rigid military-focused society and lack of individual freedoms, leading to limited economic growth and internal discord.​

One of the more glaring differences between Athens and Sparta was Sparta’s tendency toward isolation and Athens’ willingness to learn from others.  …  Athens seemed more organized overall, and able to envision the bigger pictures in preparation for potential difficulties. Athens was more diversified, nurturing many different aspects of life, setting themselves up to better handle difficulties. Sparta was concerned primarily with producing a strong military and ignored most all other aspects of society. The ability to protect yourself is important, but if there isn’t a society concerned with other aspects of life to protect, what is the point?  Sparta’s focus on its military left no opportunity to develop other aspects of society. From the reading, Society and Economy, “Throughout this period considerable wealth poured into private hands, yet Sparta was unable to make that wealth serve public purposes” (Austin, 16). These funds were not diversified, as in Athens, and therefore did nothing to help Spartan society flourish.​

Images

In the images below, there is a small watermark in the bottom right corner of the cat image; there is a small bit of glasses showing on the face of the man who is not actually wearing glasses; the text on the store image is jibberish.

You can see some of these strategies in action by looking at an image controversy from March 2024. Catherine, Princess of Wales released a photo of her with her children that several news agencies subsequently pulled, stating concerns that the image had been manipulated. You can see below digital some aspects of literacy in action, as many viewers noted inconsistencies with clothing, hair, and jagged lines. Most of these are not the kind of errors that are produced by simple photo enhancements and indicate more extreme AI involvement. This example also, unfortunately, shows what can happen when people question the reality they “see,” as numerous and often extreme conspiracy theories immediately emerged when the photo manipulation was pointed out. A better scenario would have been for the news agencies to take the time to evaluate media before releasing it. It’s also worth noting that this image has received more scrutiny than the many other manipulated images posted from influencers, etc because it was posted from an official source and displays more than minor retouching.

For more information, including image source, see here.

How to identify AI-generated images:

• Take a closer look at people, esp hands, feet, ears, noses, excess smoothing of skin. AI-generated text can appear pixelated or stretched. Similarly, if there are logos, make sure they’re the real ones and aren’t altered.​
• Check shadows and reflections for proper direction. Look for watermarks.
• Google Image Search: click on image, then click three-dot icon at right > About this image​
• aiornot.com
• Google Reverse Image Search: Upload an image to find its origin and check for duplicates. 
  • Go to Google Images (images.google.com).
  • Click the camera icon in the search bar.
  • Upload an image from your device or paste the URL of an image.
  • Review the search results to find similar images indexed by Google.

Fact-Checking Websites

• Snopes: A trusted site for debunking internet hoaxes and verifying claimsLinks to an external site.
• PolitiFact: Focuses on fact-checking political statements and claimsLinks to an external site.
• Jeffrey’s Image Metadata Viewer: Analyzes metadata of images to verify details like date, time, and locationLinks to an external site.
• Amnesty International YouTube DataViewer: Captures thumbnails of YouTube videos for reverse image searchLinks to an external site.

For both instructors and students, Critical Thinking and Media Literacy Skills must be a focus in most if not all of our courses:

• Questioning: Encouraging an inquiry-based mindset. 
• Evaluation: Assessing the credibility of sources. 
• Reflection: Considering the impact of media messages.